← All projects

Ongoing · OPmobility

Automating and levelling the weekly production plan.

A plastic injection plant, hundreds of constraints and a 2 to 4-hour meeting every week. My goal: the tool proposes and quantifies, the engineers decide.

RoleProduction engineer, project lead
IndustryAutomotive supplier, exterior plastic parts
StatusScoping done, prototype in progress
ToolsPower Query, SharePoint, Python, OR-Tools
Aerial view of the plant

01Context

The plant runs 10 injection moulding presses and more than 600 moulds. Production mixes series automotive parts, with locked delivery windows, and more flexible spare parts.

Every week, the next week’s production plan (PDP) is built from the Preactor APS, then adjusted by hand to set the number of temporary workers needed.

02Problem

  • Staffing needs were re-entered by hand every week, cross-checking 4 to 6 Excel and SharePoint sources: schedule, staffing charts, absences, ILUO skills, moulds and materials, mould deliveries.
  • These needs were not levelled: some days required far more temporary workers than others.
  • Levelling by hand took logistics and production 2 to 4 hours a week, under a very large number of constraints.

03Approach

  • Scoping. Observing the real process, mapping the data sources, separating what can be automated from what remains a human decision, defining KPIs and ROI.
  • Benchmark. APS, constraint optimisation engines, Microsoft 365 tools. Conclusion: the APS cannot level labour, and Excel alone cannot optimise.
  • Modelling. A catalogue of hard and soft constraints, and a hierarchy of objectives validated with the shop floor.

04Solution

SourcesSchedule, staffing, absences, ILUO, moulds
→
Power QueryA single data layer, no more re-entry
→
Python engineOR-Tools constraint solver
→
OutputsLevelled plan, temp staff order
  • Weekly level: the engine moves spare-part work orders to keep the number of temporary workers stable per shift, and as low as possible at night, while respecting customer dates, the arrival of external moulds, a maximum of 6 mould changes per day and as few material changes as possible.
  • Shift level: in 30-minute slots, it chains workstations and suggests job changes that respect the ILUO skills matrix.
  • A single mould reference finally connects databases that had no common identifier.

Principle: the tool proposes and quantifies, the engineers decide. Breakdowns and unexpected events are still handled on the shop floor.

05Targets

30 minlevelling meeting, down from 2 to 4 hours
0manual re-entry of staffing needs
Lessnight-shift temp work and day-to-day gaps
Fewermaterial changes (kg purged)

These are targets: they will be measured before and after on a sample of past schedules.

06Skills

Operations researchConstraint programmingData engineeringPower QuerySharePointPythonLean, SMED, load levellingProject management